holoXtrace

What is holoXtrace and what is it for?

holoXtrace is infrastructure for recording what artificial intelligence systems do in a way that is hard to alter after the fact. It is built on Holochain, so those audit records live and are checked across the people using the application rather than in a single operator's database.

holoXtrace is infrastructure for keeping tamper-proof records of what artificial intelligence systems do. It is built on Holochain, the peer-to-peer framework for applications that store their data across the machines of the people using them, so a record written by one participant is replicated and checked by others instead of being held by a single operator. The repository describes the project as audit infrastructure aimed at the transparency, accountability and human-oversight requirements of the EU AI Act.

The design divides the work across three sets of rules, called zomes, inside a Holochain application. One handles identities for the people and services doing the logging, using cryptographic keys with rotation. A second writes events into a hash chain, where each new entry carries the hash of the one before it, so changing an older record breaks the chain. A third stores provenance manifests for files and datasets in the C2PA standard, so the history attached to an asset survives attempts to strip its metadata.

One example the documentation gives is an inference event: an AI system's decision is logged together with hashes of its input and output, the model version, a confidence value and whether a human reviewed it, then linked to the previous entry in the chain. The Rust sources, the zomes and the build notes live at github.com/f13end/holoXtrace.

1 herkese açık kaynaktan derlendi; son kontrol: 23 Eylül 2026.